Level 1 moderator approach PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
 Matthew Walford posted on Wednesday, August 19, 2015 - 7:29 pm

I am hoping someone may be able to provide some guidance as I don't think my analysis is common in the literature.

Variables in data set:

One manager variable: X - is a T-Score. There is NO within group variance in X.

Sub-ordinate variables (2): Y & M - where Y is the outcome variable and M is a proposed moderator variable.

Other variables: Cluster (work teams) and XM (Interaction variable - X*M)

All variables are continuous.

It is theorised that the level 2 variable (X) should predict the Level 1 variable (Y). M should moderate this relationship.
Therefore I have a Level 1 moderator for a cross-level relationship. (M is theorised not to predict Y.)

I believe I need to adopt a means as outcome approach. From this there is a significant main effect (i.e. X-->Y). Indeed, this relationship is significant.

However, given that there is no variance at the slopes between teams, I am not sure the approach I should take in regards to modelling the moderation.

I think I need to assess how M moderates the grand slope but I am unsure of this. Would I regress the model of the main relationship X->Y onto XM?

Apologises if there are errors in my thinking.

Kind regards

 Bengt O. Muthen posted on Friday, August 21, 2015 - 1:44 pm
You could use the between part of M to moderate X -> Y on the between level. Either the observed cluster mean of M or the latent part decomposed by Mplus (see UG Chapter 9). But there isn't much going on on Within - you have M and Y but you say M doesn't predict Y. You could estimate their variances on Within.

Or, you could skip within and do a single-level analysis for between units.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message